A monte carlo comparison of bayesian estimators and trimmed means

Abstract

We indicate the usefulness of a Bayesian method of estimating a mean in the face of mean shift contamination (symmetric and asymmetric). We provide simple approximate Bayesian confidence interval formulas, and make appropriate comparisons, via Monte Carlo study, with procedures based on trimmed means, Tiku’s MML estimates, and the sample mean. We also indicate the global power of test procedures based on these estimators.

Keywords:

• Robust Estimation
• Spuriosity
• Bayesian Estimation
• Trimmed Mean
• Confidence Interval
• MML Estimators